The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions require the agent to execute a local Python script scripts/sync-threat-skill.py with specific flags (--write and --check) to synchronize and validate the threat database records.
[DYNAMIC_EXECUTION]: The update workflow relies on the execution of script-based logic to regenerate the Codex projection and validate routing scenarios, constituting dynamic behavior during the execution of the skill.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external information, such as security advisories and primary threat sources, which could contain malicious instructions intended to influence the agent's actions.
Ingestion points: Untrusted data enters the agent context via primary security sources, security advisories, and the AGENTS.md file located in external or sibling repositories.
Boundary markers: The instructions lack explicit requirements for using delimiters or providing "ignore embedded instructions" warnings when the agent processes external threat data.
Capability inventory: The agent is authorized to execute shell commands (Python 3), perform filesystem writes (via the sync script), and handle repository operations including commits, pushes, and tagging.
Sanitization: There are no documented procedures for sanitizing or escaping the content retrieved from primary sources before it is parsed or integrated into the guide.

update-threat-db